Neo Lee is co-founder of Imagine AI.
AI coding tools have already made senior engineers faster and more valuable. They can move faster by handing off the mechanical work and spending their attention where the job demands real thought. The catch is that the industry may be quietly undermining its own future. The same tools can give companies an incentive to stop hiring at the entry level, which is where the pipeline that produces senior engineers starts.
More Than Writing Code
AI is genuinely good at writing code now. New models ship every few weeks with real and steady improvements in following instructions and using tools. METR found that the length of coding jobs a frontier model can handle on its own has been doubling roughly every seven months since 2019 and closer to every four months in 2024 and 2025. Stanford’s “2026 AI Index Report” puts the productivity boost to software development at 26%. But the gains drop off quickly when it comes to work that calls for judgment.
Writing code was never the hard part of software engineering. It’s a little like writing a poem: You can string words together that sound great but still have no idea of an overarching theme. You might not see how one paragraph connects to the next or what the point of the whole piece is. In software terms, that feel for the whole is what defines good engineering, and it is the very thing that AI tools cannot yet offer.
To use a building metaphor, it’s as if we have robots now that can lay bricks and frame a house at incredible speed. But they’re also inexperienced and don’t always produce quality construction. You still need people to check the work and, more importantly, decide what kind of houses to build and where they should go: cul-de-sacs, communities, cities.
Cheap Or Expensive?
When code is cheap to generate, teams turn out more of it. Stanford research on more than 100,000 developers found that teams often feel quicker with AI early on, while at the same time racking up technical debt that drags on them later. Automated checks will find the obvious errors, but someone with experience still has to ask the harder questions. Does this change fit where the system is headed? Does it match how the product is going to evolve, or will it come back to bite the team in six months?
This creates a bottleneck at the review stage and means that the skill that matters most now is deciding what to build. Even if the “how” is now half-automated, the “what” and the “why” stay human work. To do this, you need an understanding of both the users and the system’s limits, and senior engineers have this critical skill set. Most product managers don’t have the technical depth, and AI agents lack the business or human context.
Pipeline Problems
Traditionally, a junior software engineer starts by snapping basic blocks together and then, over the years, moves up to designing how the whole structure holds together. AI automation is taking over the early rungs of the ladder, but this is where the learning happens. At the same time, some companies are looking at how AI can handle entry-level tasks and deciding they don’t need to hire junior engineers anymore. That may seem logical in the short term, but in the long term, it’s a mistake.
If a company stops hiring juniors because AI agents handle the entry-level tasks, where do its future seniors come from? Much of the learning for high-value judgment comes from getting things wrong for years and hearing about it from people smarter than you. Seniors will retire eventually, and once they’re gone, the industry faces a long-term shortage of the very judgment it’s increasingly dependent on.
The early signs of this are already showing. Juniors are being pushed into senior jobs before they’re ready, often performing manager-type work at a lower quality. Before 2023, a new grad could spend two years writing code by hand before managing one or two people. Now, they might join a startup and manage agents on day one. The role still exists, just without much or any hand-written code.
My advice for leaders is this: Don’t clear out your engineers because a headline says AI has made them obsolete. Let the weak ones go, and hold onto the strong ones, because the math points to a real shortage of senior developers on the horizon. A senior reasons about the system and the market at once, from several angles, deciding what should exist and whether it solves the right problem. Those people are easy to spot once you know the signs, so find them and keep them before the shortage arrives.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?